next up previous
Next: Overcoming Numerical Difficulties Up: Discretizations and Numerical Algorithms Previous: TR/BDF2

Nonlinear Solution Algorithms

 

A variant of Newton's method is used to solve the resulting system of nonlinear equations. Newton's method solves a system of nonlinear equations, tex2html_wrap_inline2970 , by successively applying updates in the direction of

equation1197

This involves calculation of a residual, which can be calculated directly from any of 4.24, 4.28, 4.29, and 4.30, and an approximation to the tangent ( tex2html_wrap_inline2972 ). The best approximation to the tangent is the Frechet derivative,

equation1211

For the nonlinear systems given by 4.24, 4.29, and 4.30, the tangent matrix is simply the sum of tangent of the static residual, tex2html_wrap_inline2974 , and a scaled mass matrix.

equation1221

Convergence of Newton methods can be determined using a variety of approaches. One method measures the magnitude of residual. The nonlinear method has converged if

  equation1227

or

  equation1236

Another convergence criteria measures the magnitude of the relative change in the full update. The nonlinear method has converged if

  equation1242



Dan Yergeau
Thu Nov 7 17:12:57 PST 1996